diff options
author | Vicențiu Ciorbaru <vicentiu@mariadb.org> | 2017-05-15 17:17:16 +0300 |
---|---|---|
committer | Vicențiu Ciorbaru <vicentiu@mariadb.org> | 2017-05-15 17:17:16 +0300 |
commit | 0af9818240a0745aca3fd94891664fe008de24fe (patch) | |
tree | 2074689c7de4be94ae94b13ab2d5b5831882d902 /storage/innobase | |
parent | 6ac84d9824ec384c4489b68b8087369aef147ff9 (diff) | |
download | mariadb-git-0af9818240a0745aca3fd94891664fe008de24fe.tar.gz |
5.6.36
Diffstat (limited to 'storage/innobase')
31 files changed, 658 insertions, 292 deletions
diff --git a/storage/innobase/btr/btr0sea.cc b/storage/innobase/btr/btr0sea.cc index dd28f50f4f6..01f03b2eaf7 100644 --- a/storage/innobase/btr/btr0sea.cc +++ b/storage/innobase/btr/btr0sea.cc @@ -175,7 +175,7 @@ btr_search_sys_create( btr_search_sys = (btr_search_sys_t*) mem_alloc(sizeof(btr_search_sys_t)); - btr_search_sys->hash_index = ha_create(hash_size, 0, + btr_search_sys->hash_index = ib_create(hash_size, 0, MEM_HEAP_FOR_BTR_SEARCH, 0); #if defined UNIV_AHI_DEBUG || defined UNIV_DEBUG btr_search_sys->hash_index->adaptive = TRUE; diff --git a/storage/innobase/buf/buf0buf.cc b/storage/innobase/buf/buf0buf.cc index 6f206918212..fa73e80b358 100644 --- a/storage/innobase/buf/buf0buf.cc +++ b/storage/innobase/buf/buf0buf.cc @@ -1366,7 +1366,7 @@ buf_pool_init_instance( ut_a(srv_n_page_hash_locks != 0); ut_a(srv_n_page_hash_locks <= MAX_PAGE_HASH_LOCKS); - buf_pool->page_hash = ha_create(2 * buf_pool->curr_size, + buf_pool->page_hash = ib_create(2 * buf_pool->curr_size, srv_n_page_hash_locks, MEM_HEAP_FOR_PAGE_HASH, SYNC_BUF_PAGE_HASH); diff --git a/storage/innobase/buf/buf0dblwr.cc b/storage/innobase/buf/buf0dblwr.cc index 62222993622..a752c3dfda5 100644 --- a/storage/innobase/buf/buf0dblwr.cc +++ b/storage/innobase/buf/buf0dblwr.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2014, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -355,7 +355,7 @@ recovery, this function loads the pages from double write buffer into memory. */ void buf_dblwr_init_or_load_pages( /*=========================*/ - os_file_t file, + pfs_os_file_t file, char* path, bool load_corrupt_pages) { diff --git a/storage/innobase/buf/buf0dump.cc b/storage/innobase/buf/buf0dump.cc index ed27a70307d..6c830205e2b 100644 --- a/storage/innobase/buf/buf0dump.cc +++ b/storage/innobase/buf/buf0dump.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2011, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2011, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -597,6 +597,7 @@ DECLARE_THREAD(buf_dump_thread)( void* arg MY_ATTRIBUTE((unused))) /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); ut_ad(!srv_read_only_mode); srv_buf_dump_thread_active = TRUE; @@ -632,6 +633,7 @@ DECLARE_THREAD(buf_dump_thread)( srv_buf_dump_thread_active = FALSE; + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ os_thread_exit(NULL); diff --git a/storage/innobase/buf/buf0flu.cc b/storage/innobase/buf/buf0flu.cc index 1cdd1610c4f..6550f916805 100644 --- a/storage/innobase/buf/buf0flu.cc +++ b/storage/innobase/buf/buf0flu.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -2388,6 +2388,7 @@ DECLARE_THREAD(buf_flush_page_cleaner_thread)( /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); ulint next_loop_time = ut_time_ms() + 1000; ulint n_flushed = 0; ulint last_activity = srv_get_activity_count(); @@ -2506,6 +2507,7 @@ DECLARE_THREAD(buf_flush_page_cleaner_thread)( thread_exit: buf_page_cleaner_is_active = FALSE; + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ os_thread_exit(NULL); diff --git a/storage/innobase/dict/dict0boot.cc b/storage/innobase/dict/dict0boot.cc index 1a1dd29a202..7873da30207 100644 --- a/storage/innobase/dict/dict0boot.cc +++ b/storage/innobase/dict/dict0boot.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2012, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -454,7 +454,10 @@ dict_boot(void) dberr_t err = DB_SUCCESS; - if (srv_read_only_mode && !ibuf_is_empty()) { + /** If innodb_force_recovery is set to 6 then allow + the server to start even though ibuf is not empty. */ + if (srv_force_recovery != SRV_FORCE_NO_LOG_REDO + && srv_read_only_mode && !ibuf_is_empty()) { ib_logf(IB_LOG_LEVEL_ERROR, "Change buffer must be empty when --innodb-read-only " diff --git a/storage/innobase/dict/dict0stats.cc b/storage/innobase/dict/dict0stats.cc index b0ba98308be..55a26268579 100644 --- a/storage/innobase/dict/dict0stats.cc +++ b/storage/innobase/dict/dict0stats.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2009, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2009, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -690,6 +690,9 @@ dict_stats_copy( && (src_idx = dict_table_get_next_index(src_idx)))) { if (dict_stats_should_ignore_index(dst_idx)) { + if (!(dst_idx->type & DICT_FTS)) { + dict_stats_empty_index(dst_idx); + } continue; } @@ -1096,10 +1099,10 @@ dict_stats_analyze_index_level( leaf-level delete marks because delete marks on non-leaf level do not make sense. */ - if (level == 0 && srv_stats_include_delete_marked? 0: + if (level == 0 && (srv_stats_include_delete_marked ? 0: rec_get_deleted_flag( rec, - page_is_comp(btr_pcur_get_page(&pcur)))) { + page_is_comp(btr_pcur_get_page(&pcur))))) { if (rec_is_last_on_page && !prev_rec_is_copied @@ -3229,12 +3232,6 @@ dict_stats_update( dict_table_stats_lock(table, RW_X_LATCH); - /* Initialize all stats to dummy values before - copying because dict_stats_table_clone_create() does - skip corrupted indexes so our dummy object 't' may - have less indexes than the real object 'table'. */ - dict_stats_empty_table(table); - dict_stats_copy(table, t); dict_stats_assert_initialized(table); diff --git a/storage/innobase/dict/dict0stats_bg.cc b/storage/innobase/dict/dict0stats_bg.cc index 6f01c379776..975c8a50803 100644 --- a/storage/innobase/dict/dict0stats_bg.cc +++ b/storage/innobase/dict/dict0stats_bg.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2012, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2012, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -334,6 +334,7 @@ DECLARE_THREAD(dict_stats_thread)( void* arg MY_ATTRIBUTE((unused))) /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); ut_a(!srv_read_only_mode); srv_dict_stats_thread_active = TRUE; @@ -359,6 +360,7 @@ DECLARE_THREAD(dict_stats_thread)( srv_dict_stats_thread_active = FALSE; + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit instead of return(). */ os_thread_exit(NULL); diff --git a/storage/innobase/fil/fil0fil.cc b/storage/innobase/fil/fil0fil.cc index 5e1a9d6c05e..20c6cc6201e 100644 --- a/storage/innobase/fil/fil0fil.cc +++ b/storage/innobase/fil/fil0fil.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -146,7 +146,7 @@ struct fil_node_t { belongs */ char* name; /*!< path to the file */ ibool open; /*!< TRUE if file open */ - os_file_t handle; /*!< OS handle to the file, if file open */ + pfs_os_file_t handle; /*!< OS handle to the file, if file open */ os_event_t sync_event;/*!< Condition event to group and serialize calls to fsync */ ibool is_raw_disk;/*!< TRUE if the 'file' is actually a raw @@ -316,7 +316,8 @@ initialized. */ static fil_system_t* fil_system = NULL; /** Determine if (i) is a user tablespace id or not. */ -# define fil_is_user_tablespace_id(i) ((i) > srv_undo_tablespaces_open) +# define fil_is_user_tablespace_id(i) (i != 0 \ + && !srv_is_undo_tablespace(i)) /** Determine if user has explicitly disabled fsync(). */ #ifndef __WIN__ @@ -2025,7 +2026,7 @@ UNIV_INTERN const char* fil_read_first_page( /*================*/ - os_file_t data_file, /*!< in: open data file */ + pfs_os_file_t data_file, /*!< in: open data file */ ibool one_read_already, /*!< in: TRUE if min and max parameters below already contain sensible data */ @@ -3360,7 +3361,7 @@ fil_open_linked_file( /*===============*/ const char* tablename, /*!< in: database/tablename */ char** remote_filepath,/*!< out: remote filepath */ - os_file_t* remote_file) /*!< out: remote file handle */ + pfs_os_file_t* remote_file) /*!< out: remote file handle */ { ibool success; @@ -3420,7 +3421,8 @@ fil_create_new_single_table_tablespace( tablespace file in pages, must be >= FIL_IBD_FILE_INITIAL_SIZE */ { - os_file_t file; + pfs_os_file_t file; + ibool ret; dberr_t err; byte* buf2; @@ -5862,7 +5864,7 @@ fil_flush( { fil_space_t* space; fil_node_t* node; - os_file_t file; + pfs_os_file_t file; mutex_enter(&fil_system->mutex); @@ -6224,7 +6226,7 @@ fil_buf_block_init( } struct fil_iterator_t { - os_file_t file; /*!< File handle */ + pfs_os_file_t file; /*!< File handle */ const char* filepath; /*!< File path name */ os_offset_t start; /*!< From where to start */ os_offset_t end; /*!< Where to stop */ @@ -6359,7 +6361,7 @@ fil_tablespace_iterate( PageCallback& callback) { dberr_t err; - os_file_t file; + pfs_os_file_t file; char* filepath; ut_a(n_io_buffers > 0); diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc index 41c767a4bfc..7843c328bc8 100644 --- a/storage/innobase/handler/ha_innodb.cc +++ b/storage/innobase/handler/ha_innodb.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2000, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2000, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2008, 2009 Google Inc. Copyright (c) 2009, Percona Inc. Copyright (c) 2012, Facebook Inc. @@ -5052,6 +5052,8 @@ table_opened: prebuilt->default_rec = table->s->default_values; ut_ad(prebuilt->default_rec); + prebuilt->mysql_handler = this; + /* Looks like MySQL-3.23 sometimes has primary key number != 0 */ primary_key = table->s->primary_key; key_used_on_scan = primary_key; @@ -10168,7 +10170,8 @@ ha_innobase::delete_table( extension, in contrast to ::create */ normalize_table_name(norm_name, name); - if (srv_read_only_mode) { + if (srv_read_only_mode + || srv_force_recovery >= SRV_FORCE_NO_UNDO_LOG_SCAN) { DBUG_RETURN(HA_ERR_TABLE_READONLY); } else if (row_is_magic_monitor_table(norm_name) && check_global_access(thd, PROCESS_ACL)) { diff --git a/storage/innobase/include/buf0dblwr.h b/storage/innobase/include/buf0dblwr.h index a62a6400d97..18e124b7437 100644 --- a/storage/innobase/include/buf0dblwr.h +++ b/storage/innobase/include/buf0dblwr.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2014, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -55,7 +55,7 @@ recovery, this function loads the pages from double write buffer into memory. */ void buf_dblwr_init_or_load_pages( /*=========================*/ - os_file_t file, + pfs_os_file_t file, char* path, bool load_corrupt_pages); diff --git a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h index 4d2913846b5..5806863f926 100644 --- a/storage/innobase/include/fil0fil.h +++ b/storage/innobase/include/fil0fil.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -181,7 +181,7 @@ struct fsp_open_info { ibool success; /*!< Has the tablespace been opened? */ const char* check_msg; /*!< fil_check_first_page() message */ ibool valid; /*!< Is the tablespace valid? */ - os_file_t file; /*!< File handle */ + pfs_os_file_t file; /*!< File handle */ char* filepath; /*!< File path to open */ lsn_t lsn; /*!< Flushed LSN from header page */ ulint id; /*!< Space ID */ @@ -384,7 +384,7 @@ UNIV_INTERN const char* fil_read_first_page( /*================*/ - os_file_t data_file, /*!< in: open data file */ + pfs_os_file_t data_file, /*!< in: open data file */ ibool one_read_already, /*!< in: TRUE if min and max parameters below already contain sensible data */ @@ -902,12 +902,12 @@ struct PageCallback { Called for every page in the tablespace. If the page was not updated then its state must be set to BUF_PAGE_NOT_USED. For compressed tables the page descriptor memory will be at offset: - block->frame + UNIV_PAGE_SIZE; + block->frame + UNIV_PAGE_SIZE; @param offset - physical offset within the file @param block - block read from file, note it is not from the buffer pool @retval DB_SUCCESS or error code. */ virtual dberr_t operator()( - os_offset_t offset, + os_offset_t offset, buf_block_t* block) UNIV_NOTHROW = 0; /** @@ -915,7 +915,7 @@ struct PageCallback { to open it for the file that is being iterated over. @param filename - then physical name of the tablespace file. @param file - OS file handle */ - void set_file(const char* filename, os_file_t file) UNIV_NOTHROW + void set_file(const char* filename, pfs_os_file_t file) UNIV_NOTHROW { m_file = file; m_filepath = filename; @@ -951,7 +951,7 @@ struct PageCallback { ulint m_page_size; /** File handle to the tablespace */ - os_file_t m_file; + pfs_os_file_t m_file; /** Physical file path. */ const char* m_filepath; diff --git a/storage/innobase/include/ha0ha.h b/storage/innobase/include/ha0ha.h index 7351b407e8c..58eb581e76a 100644 --- a/storage/innobase/include/ha0ha.h +++ b/storage/innobase/include/ha0ha.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1994, 2011, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1994, 2016, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -107,7 +107,7 @@ chosen to be a slightly bigger prime number. @param level in: level of the mutexes in the latching order @param n_m in: number of mutexes to protect the hash table; must be a power of 2, or 0 */ -# define ha_create(n_c,n_m,type,level) ha_create_func(n_c,level,n_m,type) +# define ib_create(n_c,n_m,type,level) ha_create_func(n_c,level,n_m,type) #else /* UNIV_SYNC_DEBUG */ /** Creates a hash table. @return own: created table @@ -116,7 +116,7 @@ chosen to be a slightly bigger prime number. @param level in: level of the mutexes in the latching order @param n_m in: number of mutexes to protect the hash table; must be a power of 2, or 0 */ -# define ha_create(n_c,n_m,type,level) ha_create_func(n_c,n_m,type) +# define ib_create(n_c,n_m,type,level) ha_create_func(n_c,n_m,type) #endif /* UNIV_SYNC_DEBUG */ /*************************************************************//** diff --git a/storage/innobase/include/os0file.h b/storage/innobase/include/os0file.h index f047302c60c..76a389f9051 100644 --- a/storage/innobase/include/os0file.h +++ b/storage/innobase/include/os0file.h @@ -1,6 +1,6 @@ /*********************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2009, Percona Inc. Portions of this file contain modifications contributed and copyrighted @@ -89,6 +89,16 @@ typedef int os_file_t; # define OS_FILE_FROM_FD(fd) fd #endif +/*Common file descriptor for file IO instrumentation with PFS +on windows and other platforms */ +struct pfs_os_file_t +{ + os_file_t m_file; +#ifdef UNIV_PFS_IO + struct PSI_file *m_psi; +#endif +}; + /** Umask for creating files */ extern ulint os_innodb_umask; @@ -216,6 +226,8 @@ extern mysql_pfs_key_t innodb_file_temp_key; various file I/O operations with performance schema. 1) register_pfs_file_open_begin() and register_pfs_file_open_end() are used to register file creation, opening, closing and renaming. +2) register_pfs_file_rename_begin() and register_pfs_file_rename_end() +are used to register file renaming 2) register_pfs_file_io_begin() and register_pfs_file_io_end() are used to register actual file read, write and flush 3) register_pfs_file_close_begin() and register_pfs_file_close_end() @@ -225,17 +237,30 @@ are used to register file deletion operations*/ do { \ locker = PSI_FILE_CALL(get_thread_file_name_locker)( \ state, key, op, name, &locker); \ - if (UNIV_LIKELY(locker != NULL)) { \ + if (locker != NULL) { \ PSI_FILE_CALL(start_file_open_wait)( \ locker, src_file, src_line); \ } \ } while (0) -# define register_pfs_file_open_end(locker, file) \ +# define register_pfs_file_open_end(locker, file, result) \ do { \ - if (UNIV_LIKELY(locker != NULL)) { \ - PSI_FILE_CALL(end_file_open_wait_and_bind_to_descriptor)(\ - locker, file); \ + if (locker != NULL) { \ + file.m_psi = PSI_FILE_CALL( \ + end_file_open_wait)( \ + locker, result); \ + } \ +} while (0) + +# define register_pfs_file_rename_begin(state, locker, key, op, name, \ + src_file, src_line) \ + register_pfs_file_open_begin(state, locker, key, op, name, \ + src_file, src_line) \ + +# define register_pfs_file_rename_end(locker, result) \ +do { \ + if (locker != NULL) { \ + PSI_FILE_CALL(end_file_open_wait)(locker, result); \ } \ } while (0) @@ -261,9 +286,9 @@ do { \ # define register_pfs_file_io_begin(state, locker, file, count, op, \ src_file, src_line) \ do { \ - locker = PSI_FILE_CALL(get_thread_file_descriptor_locker)( \ - state, file, op); \ - if (UNIV_LIKELY(locker != NULL)) { \ + locker = PSI_FILE_CALL(get_thread_file_stream_locker)( \ + state, file.m_psi, op); \ + if (locker != NULL) { \ PSI_FILE_CALL(start_file_wait)( \ locker, count, src_file, src_line); \ } \ @@ -271,7 +296,7 @@ do { \ # define register_pfs_file_io_end(locker, count) \ do { \ - if (UNIV_LIKELY(locker != NULL)) { \ + if (locker != NULL) { \ PSI_FILE_CALL(end_file_wait)(locker, count); \ } \ } while (0) @@ -289,7 +314,9 @@ os_file_rename os_aio os_file_read os_file_read_no_error_handling +os_file_read_no_error_handling_int_fd os_file_write +os_file_write_int_fd The wrapper functions have the prefix of "innodb_". */ @@ -307,7 +334,7 @@ The wrapper functions have the prefix of "innodb_". */ pfs_os_file_create_simple_no_error_handling_func( \ key, name, create_mode, access, success, __FILE__, __LINE__) -# define os_file_close(file) \ +# define os_file_close_pfs(file) \ pfs_os_file_close_func(file, __FILE__, __LINE__) # define os_aio(type, mode, name, file, buf, offset, \ @@ -315,18 +342,27 @@ The wrapper functions have the prefix of "innodb_". */ pfs_os_aio_func(type, mode, name, file, buf, offset, \ n, message1, message2, __FILE__, __LINE__) -# define os_file_read(file, buf, offset, n) \ +# define os_file_read_pfs(file, buf, offset, n) \ pfs_os_file_read_func(file, buf, offset, n, __FILE__, __LINE__) # define os_file_read_no_error_handling(file, buf, offset, n) \ pfs_os_file_read_no_error_handling_func(file, buf, offset, n, \ __FILE__, __LINE__) -# define os_file_write(name, file, buf, offset, n) \ +# define os_file_read_no_error_handling_int_fd( \ + file, buf, offset, n) \ + pfs_os_file_read_no_error_handling_int_fd_func( \ + file, buf, offset, n, __FILE__, __LINE__) + +# define os_file_write_pfs(name, file, buf, offset, n) \ pfs_os_file_write_func(name, file, buf, offset, \ n, __FILE__, __LINE__) -# define os_file_flush(file) \ +# define os_file_write_int_fd(name, file, buf, offset, n) \ + pfs_os_file_write_int_fd_func(name, file, buf, offset, \ + n, __FILE__, __LINE__) + +# define os_file_flush_pfs(file) \ pfs_os_file_flush_func(file, __FILE__, __LINE__) # define os_file_rename(key, oldpath, newpath) \ @@ -352,22 +388,29 @@ to original un-instrumented file I/O APIs */ os_file_create_simple_no_error_handling_func( \ name, create_mode, access, success) -# define os_file_close(file) os_file_close_func(file) +# define os_file_close_pfs(file) \ + os_file_close_func(file) # define os_aio(type, mode, name, file, buf, offset, n, message1, message2) \ os_aio_func(type, mode, name, file, buf, offset, n, \ message1, message2) -# define os_file_read(file, buf, offset, n) \ +# define os_file_read_pfs(file, buf, offset, n) \ os_file_read_func(file, buf, offset, n) # define os_file_read_no_error_handling(file, buf, offset, n) \ os_file_read_no_error_handling_func(file, buf, offset, n) +# define os_file_read_no_error_handling_int_fd( \ + file, buf, offset, n) \ + os_file_read_no_error_handling_func(file, buf, offset, n) -# define os_file_write(name, file, buf, offset, n) \ +# define os_file_write_int_fd(name, file, buf, offset, n) \ + os_file_write_func(name, file, buf, offset, n) +# define os_file_write_pfs(name, file, buf, offset, n) \ os_file_write_func(name, file, buf, offset, n) -# define os_file_flush(file) os_file_flush_func(file) + +# define os_file_flush_pfs(file) os_file_flush_func(file) # define os_file_rename(key, oldpath, newpath) \ os_file_rename_func(oldpath, newpath) @@ -379,6 +422,33 @@ to original un-instrumented file I/O APIs */ #endif /* UNIV_PFS_IO */ +#ifdef UNIV_PFS_IO + #define os_file_close(file) os_file_close_pfs(file) +#else + #define os_file_close(file) os_file_close_pfs((file).m_file) +#endif + +#ifdef UNIV_PFS_IO + #define os_file_read(file, buf, offset, n) \ + os_file_read_pfs(file, buf, offset, n) +#else + #define os_file_read(file, buf, offset, n) \ + os_file_read_pfs(file.m_file, buf, offset, n) +#endif + +#ifdef UNIV_PFS_IO + #define os_file_flush(file) os_file_flush_pfs(file) +#else + #define os_file_flush(file) os_file_flush_pfs(file.m_file) +#endif + +#ifdef UNIV_PFS_IO + #define os_file_write(name, file, buf, offset, n) \ + os_file_write_pfs(name, file, buf, offset, n) +#else + #define os_file_write(name, file, buf, offset, n) \ + os_file_write_pfs(name, file.m_file, buf, offset, n) +#endif /* File types for directory entry data type */ enum os_file_type_t { @@ -518,7 +588,7 @@ A simple function to open or create a file. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INTERN -os_file_t +pfs_os_file_t os_file_create_simple_no_error_handling_func( /*=========================================*/ const char* name, /*!< in: name of the file or path as a @@ -548,7 +618,7 @@ Opens an existing file or creates a new. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INTERN -os_file_t +pfs_os_file_t os_file_create_func( /*================*/ const char* name, /*!< in: name of the file or path as a @@ -617,7 +687,7 @@ os_file_create_simple() which opens or creates a file. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_simple_func( /*===========================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -640,7 +710,7 @@ monitor file creation/open. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_simple_no_error_handling_func( /*=============================================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -664,7 +734,7 @@ Add instrumentation to monitor file creation/open. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_func( /*====================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -693,7 +763,7 @@ UNIV_INLINE ibool pfs_os_file_close_func( /*===================*/ - os_file_t file, /*!< in, own: handle to a file */ + pfs_os_file_t file, /*!< in, own: handle to a file */ const char* src_file,/*!< in: file name where func invoked */ ulint src_line);/*!< in: line where the func invoked */ /*******************************************************************//** @@ -706,7 +776,7 @@ UNIV_INLINE ibool pfs_os_file_read_func( /*==================*/ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read */ os_offset_t offset, /*!< in: file offset where to read */ ulint n, /*!< in: number of bytes to read */ @@ -724,7 +794,7 @@ UNIV_INLINE ibool pfs_os_file_read_no_error_handling_func( /*====================================*/ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read */ os_offset_t offset, /*!< in: file offset where to read */ ulint n, /*!< in: number of bytes to read */ @@ -745,7 +815,7 @@ pfs_os_aio_func( ulint mode, /*!< in: OS_AIO_NORMAL etc. I/O mode */ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read or from which to write */ os_offset_t offset, /*!< in: file offset where to read or write */ @@ -772,7 +842,7 @@ pfs_os_file_write_func( /*===================*/ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ const void* buf, /*!< in: buffer from which to write */ os_offset_t offset, /*!< in: file offset where to write */ ulint n, /*!< in: number of bytes to write */ @@ -789,7 +859,7 @@ UNIV_INLINE ibool pfs_os_file_flush_func( /*===================*/ - os_file_t file, /*!< in, own: handle to a file */ + pfs_os_file_t file, /*!< in, own: handle to a file */ const char* src_file,/*!< in: file name where func invoked */ ulint src_line);/*!< in: line where the func invoked */ @@ -860,7 +930,7 @@ UNIV_INTERN os_offset_t os_file_get_size( /*=============*/ - os_file_t file) /*!< in: handle to a file */ + pfs_os_file_t file) /*!< in: handle to a file */ MY_ATTRIBUTE((warn_unused_result)); /***********************************************************************//** Write the specified number of zeros to a newly created file. @@ -871,7 +941,7 @@ os_file_set_size( /*=============*/ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ os_offset_t size) /*!< in: file size */ MY_ATTRIBUTE((nonnull, warn_unused_result)); /***********************************************************************//** @@ -1109,7 +1179,7 @@ os_aio_func( caution! */ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read or from which to write */ os_offset_t offset, /*!< in: file offset where to read or write */ diff --git a/storage/innobase/include/os0file.ic b/storage/innobase/include/os0file.ic index defd8204ba3..31d5cc1d6a2 100644 --- a/storage/innobase/include/os0file.ic +++ b/storage/innobase/include/os0file.ic @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 2010, 2011, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 2010, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -34,7 +34,7 @@ os_file_create_simple() which opens or creates a file. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_simple_func( /*===========================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -47,7 +47,7 @@ pfs_os_file_create_simple_func( const char* src_file,/*!< in: file name where func invoked */ ulint src_line)/*!< in: line where the func invoked */ { - os_file_t file; + pfs_os_file_t file; struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; @@ -58,11 +58,13 @@ pfs_os_file_create_simple_func( : PSI_FILE_OPEN), name, src_file, src_line); - file = os_file_create_simple_func(name, create_mode, + file.m_file = os_file_create_simple_func(name, create_mode, access_type, success); + file.m_psi = NULL; - /* Regsiter the returning "file" value with the system */ - register_pfs_file_open_end(locker, file); + /* Regsiter psi value for the file */ + register_pfs_file_open_end(locker, file, + (*success == TRUE ? success : 0)); return(file); } @@ -76,7 +78,7 @@ monitor file creation/open. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_simple_no_error_handling_func( /*=============================================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -91,7 +93,7 @@ pfs_os_file_create_simple_no_error_handling_func( const char* src_file,/*!< in: file name where func invoked */ ulint src_line)/*!< in: line where the func invoked */ { - os_file_t file; + pfs_os_file_t file; struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; @@ -104,8 +106,10 @@ pfs_os_file_create_simple_no_error_handling_func( file = os_file_create_simple_no_error_handling_func( name, create_mode, access_type, success); + file.m_psi = NULL; - register_pfs_file_open_end(locker, file); + register_pfs_file_open_end(locker, file, + (*success == TRUE ? success : 0)); return(file); } @@ -118,7 +122,7 @@ Add instrumentation to monitor file creation/open. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INLINE -os_file_t +pfs_os_file_t pfs_os_file_create_func( /*====================*/ mysql_pfs_key_t key, /*!< in: Performance Schema Key */ @@ -137,7 +141,7 @@ pfs_os_file_create_func( const char* src_file,/*!< in: file name where func invoked */ ulint src_line)/*!< in: line where the func invoked */ { - os_file_t file; + pfs_os_file_t file; struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; @@ -149,8 +153,10 @@ pfs_os_file_create_func( name, src_file, src_line); file = os_file_create_func(name, create_mode, purpose, type, success); + file.m_psi = NULL; - register_pfs_file_open_end(locker, file); + register_pfs_file_open_end(locker, file, + (*success == TRUE ? success : 0)); return(file); } @@ -164,7 +170,7 @@ UNIV_INLINE ibool pfs_os_file_close_func( /*===================*/ - os_file_t file, /*!< in, own: handle to a file */ + pfs_os_file_t file, /*!< in, own: handle to a file */ const char* src_file,/*!< in: file name where func invoked */ ulint src_line)/*!< in: line where the func invoked */ { @@ -176,7 +182,7 @@ pfs_os_file_close_func( register_pfs_file_io_begin(&state, locker, file, 0, PSI_FILE_CLOSE, src_file, src_line); - result = os_file_close_func(file); + result = os_file_close_func(file.m_file); register_pfs_file_io_end(locker, 0); @@ -197,7 +203,7 @@ pfs_os_aio_func( ulint mode, /*!< in: OS_AIO_NORMAL etc. I/O mode */ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read or from which to write */ os_offset_t offset, /*!< in: file offset where to read or write */ @@ -242,7 +248,7 @@ UNIV_INLINE ibool pfs_os_file_read_func( /*==================*/ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read */ os_offset_t offset, /*!< in: file offset where to read */ ulint n, /*!< in: number of bytes to read */ @@ -256,7 +262,7 @@ pfs_os_file_read_func( register_pfs_file_io_begin(&state, locker, file, n, PSI_FILE_READ, src_file, src_line); - result = os_file_read_func(file, buf, offset, n); + result = os_file_read_func(file.m_file, buf, offset, n); register_pfs_file_io_end(locker, n); @@ -275,7 +281,7 @@ UNIV_INLINE ibool pfs_os_file_read_no_error_handling_func( /*====================================*/ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read */ os_offset_t offset, /*!< in: file offset where to read */ ulint n, /*!< in: number of bytes to read */ @@ -289,13 +295,50 @@ pfs_os_file_read_no_error_handling_func( register_pfs_file_io_begin(&state, locker, file, n, PSI_FILE_READ, src_file, src_line); - result = os_file_read_no_error_handling_func(file, buf, offset, n); + result = os_file_read_no_error_handling_func(file.m_file, buf, offset, n); register_pfs_file_io_end(locker, n); return(result); } +/** NOTE! Please use the corresponding macro +os_file_read_no_error_handling_int_fd(), not directly this function! +This is the performance schema instrumented wrapper function for +os_file_read_no_error_handling_int_fd_func() which requests a +synchronous read operation. +@return TRUE if request was successful, FALSE if fail */ +UNIV_INLINE +ibool +pfs_os_file_read_no_error_handling_int_fd_func( + int file, /*!< in: handle to a file */ + void* buf, /*!< in: buffer where to read */ + os_offset_t offset, /*!< in: file offset where to read */ + ulint n, /*!< in: number of bytes to read */ + const char* src_file,/*!< in: file name where func invoked */ + ulint src_line)/*!< in: line where the func invoked */ +{ + + PSI_file_locker_state state; + struct PSI_file_locker* locker = NULL; + + locker = PSI_FILE_CALL(get_thread_file_descriptor_locker)( + &state, file, PSI_FILE_READ); + if (locker != NULL) { + PSI_FILE_CALL(start_file_wait)( + locker, n, + __FILE__, __LINE__); + } + ibool result = os_file_read_no_error_handling_func( + OS_FILE_FROM_FD(file), buf, offset, n); + + if (locker != NULL) { + PSI_FILE_CALL(end_file_wait)(locker, n); + } + + return(result); +} + /*******************************************************************//** NOTE! Please use the corresponding macro os_file_write(), not directly this function! @@ -308,7 +351,7 @@ pfs_os_file_write_func( /*===================*/ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ const void* buf, /*!< in: buffer from which to write */ os_offset_t offset, /*!< in: file offset where to write */ ulint n, /*!< in: number of bytes to write */ @@ -322,13 +365,50 @@ pfs_os_file_write_func( register_pfs_file_io_begin(&state, locker, file, n, PSI_FILE_WRITE, src_file, src_line); - result = os_file_write_func(name, file, buf, offset, n); + result = os_file_write_func(name, file.m_file, buf, offset, n); register_pfs_file_io_end(locker, n); return(result); } +/** NOTE! Please use the corresponding macro os_file_write(), not +directly this function! +This is the performance schema instrumented wrapper function for +os_file_write() which requests a synchronous write operation. +@return TRUE if request was successful, FALSE if fail */ +UNIV_INLINE +ibool +pfs_os_file_write_int_fd_func( + const char* name, /*!< in: name of the file or path as a + null-terminated string */ + int file, /*!< in: handle to a file */ + const void* buf, /*!< in: buffer from which to write */ + os_offset_t offset, /*!< in: file offset where to write */ + ulint n, /*!< in: number of bytes to write */ + const char* src_file,/*!< in: file name where func invoked */ + ulint src_line)/*!< in: line where the func invoked */ +{ + PSI_file_locker_state state; + struct PSI_file_locker* locker = NULL; + + locker = PSI_FILE_CALL(get_thread_file_descriptor_locker)( + &state, file, PSI_FILE_WRITE); + if (locker != NULL) { + PSI_FILE_CALL(start_file_wait)( + locker, n, + __FILE__, __LINE__); + } + ibool result = os_file_write_func( + name, OS_FILE_FROM_FD(file), buf, offset, n); + + if (locker != NULL) { + PSI_FILE_CALL(end_file_wait)(locker, n); + } + + return(result); +} + /***********************************************************************//** NOTE! Please use the corresponding macro os_file_flush(), not directly this function! @@ -339,7 +419,7 @@ UNIV_INLINE ibool pfs_os_file_flush_func( /*===================*/ - os_file_t file, /*!< in, own: handle to a file */ + pfs_os_file_t file, /*!< in, own: handle to a file */ const char* src_file,/*!< in: file name where func invoked */ ulint src_line)/*!< in: line where the func invoked */ { @@ -349,7 +429,7 @@ pfs_os_file_flush_func( register_pfs_file_io_begin(&state, locker, file, 0, PSI_FILE_SYNC, src_file, src_line); - result = os_file_flush_func(file); + result = os_file_flush_func(file.m_file); register_pfs_file_io_end(locker, 0); @@ -377,12 +457,12 @@ pfs_os_file_rename_func( struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; - register_pfs_file_open_begin(&state, locker, key, PSI_FILE_RENAME, newpath, + register_pfs_file_rename_begin(&state, locker, key, PSI_FILE_RENAME, newpath, src_file, src_line); result = os_file_rename_func(oldpath, newpath); - register_pfs_file_open_end(locker, 0); + register_pfs_file_rename_end(locker, 0); return(result); } diff --git a/storage/innobase/include/row0mysql.h b/storage/innobase/include/row0mysql.h index fc1846b76f3..d0c1bdaa585 100644 --- a/storage/innobase/include/row0mysql.h +++ b/storage/innobase/include/row0mysql.h @@ -652,6 +652,8 @@ struct mysql_row_templ_t { #define ROW_PREBUILT_ALLOCATED 78540783 #define ROW_PREBUILT_FREED 26423527 +class ha_innobase; + /** A struct for (sometimes lazily) prebuilt structures in an Innobase table handle used within MySQL; these are used to save CPU time. */ @@ -879,6 +881,12 @@ struct row_prebuilt_t { to InnoDB format.*/ uint srch_key_val_len; /*!< Size of search key */ + /** MySQL handler object. */ + ha_innobase* mysql_handler; + + /** True if exceeded the end_range while filling the prefetch cache. */ + bool end_range; + }; /** Callback for row_mysql_sys_index_iterate() */ diff --git a/storage/innobase/include/srv0srv.h b/storage/innobase/include/srv0srv.h index 042ced75d10..7d701536341 100644 --- a/storage/innobase/include/srv0srv.h +++ b/storage/innobase/include/srv0srv.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All rights reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All rights reserved. Copyright (c) 2008, 2009, Google Inc. Copyright (c) 2009, Percona Inc. @@ -785,6 +785,13 @@ void srv_purge_wakeup(void); /*==================*/ +/** Check whether given space id is undo tablespace id +@param[in] space_id space id to check +@return true if it is undo tablespace else false. */ +bool +srv_is_undo_tablespace( + ulint space_id); + /** Status variables to be passed to MySQL */ struct export_var_t{ ulint innodb_data_pending_reads; /*!< Pending reads */ diff --git a/storage/innobase/include/srv0start.h b/storage/innobase/include/srv0start.h index 963b767f0fb..a60776a4665 100644 --- a/storage/innobase/include/srv0start.h +++ b/storage/innobase/include/srv0start.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -139,6 +139,8 @@ extern ibool srv_startup_is_before_trx_rollback_phase; /** TRUE if a raw partition is in use */ extern ibool srv_start_raw_disk_in_use; +/** Undo tablespaces starts with space_id. */ +extern ulint srv_undo_space_id_start; /** Shutdown state */ enum srv_shutdown_state { diff --git a/storage/innobase/include/trx0xa.h b/storage/innobase/include/trx0xa.h index 7caddfb7ba4..255431293f5 100644 --- a/storage/innobase/include/trx0xa.h +++ b/storage/innobase/include/trx0xa.h @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2009, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -24,6 +24,8 @@ this program; if not, write to the Free Software Foundation, Inc., #ifndef XA_H #define XA_H +#include "xa.h" + /* * Transaction branch identification: XID and NULLXID: */ @@ -35,17 +37,6 @@ this program; if not, write to the Free Software Foundation, Inc., #define MAXGTRIDSIZE 64 /*!< maximum size in bytes of gtrid */ #define MAXBQUALSIZE 64 /*!< maximum size in bytes of bqual */ -/** X/Open XA distributed transaction identifier */ -struct xid_t { - long formatID; /*!< format identifier; -1 - means that the XID is null */ - long gtrid_length; /*!< value from 1 through 64 */ - long bqual_length; /*!< value from 1 through 64 */ - char data[XIDDATASIZE]; /*!< distributed transaction - identifier */ -}; -/** X/Open XA distributed transaction identifier */ -typedef struct xid_t XID; #endif /** X/Open XA distributed transaction status codes */ /* @{ */ diff --git a/storage/innobase/include/univ.i b/storage/innobase/include/univ.i index dfb59673f0b..5d3dcafb322 100644 --- a/storage/innobase/include/univ.i +++ b/storage/innobase/include/univ.i @@ -133,14 +133,8 @@ HAVE_PSI_INTERFACE is defined. */ #if defined HAVE_PSI_INTERFACE && !defined UNIV_HOTBACKUP # define UNIV_PFS_MUTEX # define UNIV_PFS_RWLOCK -/* For I/O instrumentation, performance schema rely -on a native descriptor to identify the file, this -descriptor could conflict with our OS level descriptor. -Disable IO instrumentation on Windows until this is -resolved */ -# ifndef __WIN__ -# define UNIV_PFS_IO -# endif + +# define UNIV_PFS_IO # define UNIV_PFS_THREAD /* There are mutexes/rwlocks that we want to exclude from diff --git a/storage/innobase/log/log0log.cc b/storage/innobase/log/log0log.cc index 8ff8e39d352..51b057ec09a 100644 --- a/storage/innobase/log/log0log.cc +++ b/storage/innobase/log/log0log.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2009, Google Inc. Portions of this file contain modifications contributed and copyrighted by @@ -3406,7 +3406,12 @@ loop: lsn = log_sys->lsn; - if (lsn != log_sys->last_checkpoint_lsn + const bool is_last = ((srv_force_recovery == SRV_FORCE_NO_LOG_REDO + && lsn == log_sys->last_checkpoint_lsn + + LOG_BLOCK_HDR_SIZE) + || lsn == log_sys->last_checkpoint_lsn); + + if (!is_last #ifdef UNIV_LOG_ARCHIVE || (srv_log_archive_on && lsn != log_sys->archived_lsn + LOG_BLOCK_HDR_SIZE) diff --git a/storage/innobase/log/log0recv.cc b/storage/innobase/log/log0recv.cc index 85f4f6ea671..69d74a99ab5 100644 --- a/storage/innobase/log/log0recv.cc +++ b/storage/innobase/log/log0recv.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1997, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1997, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2012, Facebook Inc. This program is free software; you can redistribute it and/or modify it under @@ -332,6 +332,7 @@ DECLARE_THREAD(recv_writer_thread)( /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); ut_ad(!srv_read_only_mode); #ifdef UNIV_PFS_THREAD @@ -364,6 +365,7 @@ DECLARE_THREAD(recv_writer_thread)( recv_writer_thread_active = false; + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ diff --git a/storage/innobase/os/os0file.cc b/storage/innobase/os/os0file.cc index 46a7b801521..dbd30c3d658 100644 --- a/storage/innobase/os/os0file.cc +++ b/storage/innobase/os/os0file.cc @@ -1,6 +1,6 @@ /*********************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2009, Percona Inc. Portions of this file contain modifications contributed and copyrighted @@ -163,7 +163,7 @@ struct os_aio_slot_t{ byte* buf; /*!< buffer used in i/o */ ulint type; /*!< OS_FILE_READ or OS_FILE_WRITE */ os_offset_t offset; /*!< file offset in bytes */ - os_file_t file; /*!< file where to read or write */ + pfs_os_file_t file; /*!< file where to read or write */ const char* name; /*!< file name or path */ ibool io_already_done;/*!< used only in simulated aio: TRUE if the physical i/o already @@ -1306,7 +1306,7 @@ A simple function to open or create a file. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INTERN -os_file_t +pfs_os_file_t os_file_create_simple_no_error_handling_func( /*=========================================*/ const char* name, /*!< in: name of the file or path as a @@ -1318,7 +1318,7 @@ os_file_create_simple_no_error_handling_func( used by a backup program reading the file */ ibool* success)/*!< out: TRUE if succeed, FALSE if error */ { - os_file_t file; + pfs_os_file_t file; *success = FALSE; #ifdef __WIN__ @@ -1326,7 +1326,6 @@ os_file_create_simple_no_error_handling_func( DWORD create_flag; DWORD attributes = 0; DWORD share_mode = FILE_SHARE_READ; - ut_a(name); ut_a(!(create_mode & OS_FILE_ON_ERROR_SILENT)); @@ -1343,8 +1342,8 @@ os_file_create_simple_no_error_handling_func( ib_logf(IB_LOG_LEVEL_ERROR, "Unknown file create mode (%lu) for file '%s'", create_mode, name); - - return((os_file_t) -1); + file.m_file = (os_file_t)-1; + return(file); } if (access_type == OS_FILE_READ_ONLY) { @@ -1367,11 +1366,11 @@ os_file_create_simple_no_error_handling_func( ib_logf(IB_LOG_LEVEL_ERROR, "Unknown file access type (%lu) for file '%s'", access_type, name); - - return((os_file_t) -1); + file.m_file = (os_file_t)-1; + return(file); } - file = CreateFile((LPCTSTR) name, + file.m_file = CreateFile((LPCTSTR) name, access, share_mode, NULL, // Security attributes @@ -1379,11 +1378,10 @@ os_file_create_simple_no_error_handling_func( attributes, NULL); // No template file - *success = (file != INVALID_HANDLE_VALUE); + *success = (file.m_file != INVALID_HANDLE_VALUE); #else /* __WIN__ */ int create_flag; const char* mode_str = NULL; - ut_a(name); ut_a(!(create_mode & OS_FILE_ON_ERROR_SILENT)); @@ -1425,13 +1423,13 @@ os_file_create_simple_no_error_handling_func( ib_logf(IB_LOG_LEVEL_ERROR, "Unknown file create mode (%lu) for file '%s'", create_mode, name); - - return((os_file_t) -1); + file.m_file = -1; + return(file); } - file = ::open(name, create_flag, os_innodb_umask); + file.m_file = ::open(name, create_flag, os_innodb_umask); - *success = file == -1 ? FALSE : TRUE; + *success = file.m_file == -1 ? FALSE : TRUE; /* This function is always called for data files, we should disable OS caching (O_DIRECT) here as we do in os_file_create_func(), so @@ -1441,18 +1439,18 @@ os_file_create_simple_no_error_handling_func( && (srv_unix_file_flush_method == SRV_UNIX_O_DIRECT || srv_unix_file_flush_method == SRV_UNIX_O_DIRECT_NO_FSYNC)) { - os_file_set_nocache(file, name, mode_str); + os_file_set_nocache(file.m_file, name, mode_str); } #ifdef USE_FILE_LOCK if (!srv_read_only_mode && *success && access_type == OS_FILE_READ_WRITE - && os_file_lock(file, name)) { + && os_file_lock(file.m_file, name)) { *success = FALSE; - close(file); - file = -1; + close(file.m_file); + file.m_file = -1; } #endif /* USE_FILE_LOCK */ @@ -1527,7 +1525,7 @@ Opens an existing file or creates a new. @return own: handle to the file, not defined if error, error number can be retrieved with os_file_get_last_error */ UNIV_INTERN -os_file_t +pfs_os_file_t os_file_create_func( /*================*/ const char* name, /*!< in: name of the file or path as a @@ -1543,24 +1541,25 @@ os_file_create_func( ulint type, /*!< in: OS_DATA_FILE or OS_LOG_FILE */ ibool* success)/*!< out: TRUE if succeed, FALSE if error */ { - os_file_t file; + pfs_os_file_t file; ibool retry; ibool on_error_no_exit; ibool on_error_silent; - #ifdef __WIN__ DBUG_EXECUTE_IF( "ib_create_table_fail_disk_full", *success = FALSE; SetLastError(ERROR_DISK_FULL); - return((os_file_t) -1); + file.m_file = (os_file_t)-1; + return(file); ); #else /* __WIN__ */ DBUG_EXECUTE_IF( "ib_create_table_fail_disk_full", *success = FALSE; errno = ENOSPC; - return((os_file_t) -1); + file.m_file = -1; + return(file); ); #endif /* __WIN__ */ @@ -1611,7 +1610,8 @@ os_file_create_func( "Unknown file create mode (%lu) for file '%s'", create_mode, name); - return((os_file_t) -1); + file.m_file = (os_file_t)-1; + return(file); } DWORD attributes = 0; @@ -1636,8 +1636,8 @@ os_file_create_func( ib_logf(IB_LOG_LEVEL_ERROR, "Unknown purpose flag (%lu) while opening file '%s'", purpose, name); - - return((os_file_t)(-1)); + file.m_file = (os_file_t)-1; + return(file); } #ifdef UNIV_NON_BUFFERED_IO @@ -1664,11 +1664,11 @@ os_file_create_func( do { /* Use default security attributes and no template file. */ - file = CreateFile( + file.m_file = CreateFile( (LPCTSTR) name, access, share_mode, NULL, create_flag, attributes, NULL); - if (file == INVALID_HANDLE_VALUE) { + if (file.m_file == INVALID_HANDLE_VALUE) { const char* operation; operation = (create_mode == OS_FILE_CREATE @@ -1693,7 +1693,6 @@ os_file_create_func( #else /* __WIN__ */ int create_flag; const char* mode_str = NULL; - on_error_no_exit = create_mode & OS_FILE_ON_ERROR_NO_EXIT ? TRUE : FALSE; on_error_silent = create_mode & OS_FILE_ON_ERROR_SILENT @@ -1731,7 +1730,8 @@ os_file_create_func( "Unknown file create mode (%lu) for file '%s'", create_mode, name); - return((os_file_t) -1); + file.m_file = -1; + return(file); } ut_a(type == OS_LOG_FILE || type == OS_DATA_FILE); @@ -1751,9 +1751,9 @@ os_file_create_func( #endif /* O_SYNC */ do { - file = ::open(name, create_flag, os_innodb_umask); + file.m_file = ::open(name, create_flag, os_innodb_umask); - if (file == -1) { + if (file.m_file == -1) { const char* operation; operation = (create_mode == OS_FILE_CREATE @@ -1783,14 +1783,14 @@ os_file_create_func( && (srv_unix_file_flush_method == SRV_UNIX_O_DIRECT || srv_unix_file_flush_method == SRV_UNIX_O_DIRECT_NO_FSYNC)) { - os_file_set_nocache(file, name, mode_str); + os_file_set_nocache(file.m_file, name, mode_str); } #ifdef USE_FILE_LOCK if (!srv_read_only_mode && *success && create_mode != OS_FILE_OPEN_RAW - && os_file_lock(file, name)) { + && os_file_lock(file.m_file, name)) { if (create_mode == OS_FILE_OPEN_RETRY) { @@ -1802,7 +1802,7 @@ os_file_create_func( for (int i = 0; i < 100; i++) { os_thread_sleep(1000000); - if (!os_file_lock(file, name)) { + if (!os_file_lock(file.m_file, name)) { *success = TRUE; return(file); } @@ -1813,8 +1813,8 @@ os_file_create_func( } *success = FALSE; - close(file); - file = -1; + close(file.m_file); + file.m_file = -1; } #endif /* USE_FILE_LOCK */ @@ -2086,14 +2086,14 @@ UNIV_INTERN os_offset_t os_file_get_size( /*=============*/ - os_file_t file) /*!< in: handle to a file */ + pfs_os_file_t file) /*!< in: handle to a file */ { #ifdef __WIN__ os_offset_t offset; DWORD high; DWORD low; - low = GetFileSize(file, &high); + low = GetFileSize(file.m_file, &high); if ((low == 0xFFFFFFFF) && (GetLastError() != NO_ERROR)) { return((os_offset_t) -1); @@ -2103,7 +2103,8 @@ os_file_get_size( return(offset); #else - return((os_offset_t) lseek(file, 0, SEEK_END)); + return((os_offset_t) lseek(file.m_file, 0, SEEK_END)); + #endif /* __WIN__ */ } @@ -2116,7 +2117,7 @@ os_file_set_size( /*=============*/ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ os_offset_t size) /*!< in: file size */ { os_offset_t current_size; @@ -4187,7 +4188,7 @@ os_aio_array_reserve_slot( the aio operation */ void* message2,/*!< in: message to be passed along with the aio operation */ - os_file_t file, /*!< in: file handle */ + pfs_os_file_t file, /*!< in: file handle */ const char* name, /*!< in: name of the file or path as a null-terminated string */ void* buf, /*!< in: buffer where to read or from which @@ -4307,10 +4308,10 @@ found: iocb = &slot->control; if (type == OS_FILE_READ) { - io_prep_pread(iocb, file, buf, len, aio_offset); + io_prep_pread(iocb, file.m_file, buf, len, aio_offset); } else { ut_a(type == OS_FILE_WRITE); - io_prep_pwrite(iocb, file, buf, len, aio_offset); + io_prep_pwrite(iocb, file.m_file, buf, len, aio_offset); } iocb->data = (void*) slot; @@ -4548,7 +4549,7 @@ os_aio_func( caution! */ const char* name, /*!< in: name of the file or path as a null-terminated string */ - os_file_t file, /*!< in: handle to a file */ + pfs_os_file_t file, /*!< in: handle to a file */ void* buf, /*!< in: buffer where to read or from which to write */ os_offset_t offset, /*!< in: file offset where to read or write */ @@ -4573,8 +4574,7 @@ os_aio_func( ulint dummy_type; #endif /* WIN_ASYNC_IO */ ulint wake_later; - - ut_ad(file); + ut_ad(file.m_file); ut_ad(buf); ut_ad(n > 0); ut_ad(n % OS_FILE_LOG_BLOCK_SIZE == 0); @@ -4606,13 +4606,11 @@ os_aio_func( and os_file_write_func() */ if (type == OS_FILE_READ) { - return(os_file_read_func(file, buf, offset, n)); + return(os_file_read_func(file.m_file, buf, offset, n)); } - ut_ad(!srv_read_only_mode); ut_a(type == OS_FILE_WRITE); - - return(os_file_write_func(name, file, buf, offset, n)); + return(os_file_write_func(name, file.m_file, buf, offset, n)); } try_again: @@ -4664,9 +4662,8 @@ try_again: os_n_file_reads++; os_bytes_read_since_printout += n; #ifdef WIN_ASYNC_IO - ret = ReadFile(file, buf, (DWORD) n, &len, + ret = ReadFile(file.m_file, buf, (DWORD) n, &len, &(slot->control)); - #elif defined(LINUX_NATIVE_AIO) if (!os_aio_linux_dispatch(array, slot)) { goto err_exit; @@ -4684,9 +4681,8 @@ try_again: if (srv_use_native_aio) { os_n_file_writes++; #ifdef WIN_ASYNC_IO - ret = WriteFile(file, buf, (DWORD) n, &len, + ret = WriteFile(file.m_file, buf, (DWORD) n, &len, &(slot->control)); - #elif defined(LINUX_NATIVE_AIO) if (!os_aio_linux_dispatch(array, slot)) { goto err_exit; @@ -4840,8 +4836,7 @@ os_aio_windows_handle( srv_set_io_thread_op_info( orig_seg, "get windows aio return value"); } - - ret = GetOverlappedResult(slot->file, &(slot->control), &len, TRUE); + ret = GetOverlappedResult(slot->file.m_file, &(slot->control), &len, TRUE); *message1 = slot->message1; *message2 = slot->message2; @@ -4870,7 +4865,8 @@ os_aio_windows_handle( and os_file_write APIs, need to register with performance schema explicitly here. */ struct PSI_file_locker* locker = NULL; - register_pfs_file_io_begin(locker, slot->file, slot->len, + PSI_file_locker_state state; + register_pfs_file_io_begin(&state, locker, slot->file, slot->len, (slot->type == OS_FILE_WRITE) ? PSI_FILE_WRITE : PSI_FILE_READ, @@ -4881,16 +4877,14 @@ os_aio_windows_handle( switch (slot->type) { case OS_FILE_WRITE: - ret = WriteFile(slot->file, slot->buf, + ret = WriteFile(slot->file.m_file, slot->buf, (DWORD) slot->len, &len, &(slot->control)); - break; case OS_FILE_READ: - ret = ReadFile(slot->file, slot->buf, + ret = ReadFile(slot->file.m_file, slot->buf, (DWORD) slot->len, &len, &(slot->control)); - break; default: ut_error; @@ -4906,8 +4900,7 @@ os_aio_windows_handle( file where we also use async i/o: in Windows we must use the same wait mechanism as for async i/o */ - - ret = GetOverlappedResult(slot->file, + ret = GetOverlappedResult(slot->file.m_file, &(slot->control), &len, TRUE); } @@ -5354,12 +5347,11 @@ consecutive_loop: os_aio_slot_t* slot; slot = os_aio_array_get_nth_slot(array, i + segment * n); - if (slot->reserved && slot != aio_slot && slot->offset == aio_slot->offset + aio_slot->len && slot->type == aio_slot->type - && slot->file == aio_slot->file) { + && slot->file.m_file == aio_slot->file.m_file) { /* Found a consecutive i/o request */ diff --git a/storage/innobase/row/row0log.cc b/storage/innobase/row/row0log.cc index 54183759e8d..4596e2fb951 100644 --- a/storage/innobase/row/row0log.cc +++ b/storage/innobase/row/row0log.cc @@ -363,9 +363,9 @@ row_log_online_op( goto err_exit; } - ret = os_file_write( + ret = os_file_write_int_fd( "(modification log)", - OS_FILE_FROM_FD(log->fd), + log->fd, log->tail.block, byte_offset, srv_sort_buf_size); log->tail.blocks++; if (!ret) { @@ -479,9 +479,9 @@ row_log_table_close_func( goto err_exit; } - ret = os_file_write( + ret = os_file_write_int_fd( "(modification log)", - OS_FILE_FROM_FD(log->fd), + log->fd, log->tail.block, byte_offset, srv_sort_buf_size); log->tail.blocks++; if (!ret) { @@ -2609,11 +2609,10 @@ all_done: goto func_exit; } - success = os_file_read_no_error_handling( - OS_FILE_FROM_FD(index->online_log->fd), + success = os_file_read_no_error_handling_int_fd( + index->online_log->fd, index->online_log->head.block, ofs, srv_sort_buf_size); - if (!success) { fprintf(stderr, "InnoDB: unable to read temporary file" " for table %s\n", index->table_name); @@ -3436,8 +3435,8 @@ all_done: goto func_exit; } - success = os_file_read_no_error_handling( - OS_FILE_FROM_FD(index->online_log->fd), + success = os_file_read_no_error_handling_int_fd( + index->online_log->fd, index->online_log->head.block, ofs, srv_sort_buf_size); diff --git a/storage/innobase/row/row0merge.cc b/storage/innobase/row/row0merge.cc index f8bea67906c..2d63352feaf 100644 --- a/storage/innobase/row/row0merge.cc +++ b/storage/innobase/row/row0merge.cc @@ -869,8 +869,9 @@ row_merge_read( } #endif /* UNIV_DEBUG */ - success = os_file_read_no_error_handling(OS_FILE_FROM_FD(fd), buf, + success = os_file_read_no_error_handling_int_fd(fd, buf, ofs, srv_sort_buf_size); + #ifdef POSIX_FADV_DONTNEED /* Each block is read exactly once. Free up the file cache. */ posix_fadvise(fd, ofs, srv_sort_buf_size, POSIX_FADV_DONTNEED); @@ -904,7 +905,7 @@ row_merge_write( DBUG_EXECUTE_IF("row_merge_write_failure", return(FALSE);); - ret = os_file_write("(merge)", OS_FILE_FROM_FD(fd), buf, ofs, buf_len); + ret = os_file_write_int_fd("(merge)", fd, buf, ofs, buf_len); #ifdef UNIV_DEBUG if (row_merge_print_block_write) { @@ -3121,14 +3122,21 @@ row_merge_file_create_low( performance schema */ struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; - register_pfs_file_open_begin(&state, locker, innodb_file_temp_key, - PSI_FILE_OPEN, - "Innodb Merge Temp File", - __FILE__, __LINE__); + locker = PSI_FILE_CALL(get_thread_file_name_locker)( + &state, innodb_file_temp_key, PSI_FILE_OPEN, + "Innodb Merge Temp File", &locker); + if (locker != NULL) { + PSI_FILE_CALL(start_file_open_wait)(locker, + __FILE__, + __LINE__); + } #endif fd = innobase_mysql_tmpfile(path); #ifdef UNIV_PFS_IO - register_pfs_file_open_end(locker, fd); + if (locker != NULL) { + PSI_FILE_CALL(end_file_open_wait_and_bind_to_descriptor)( + locker, fd); + } #endif if (fd < 0) { @@ -3175,15 +3183,20 @@ row_merge_file_destroy_low( #ifdef UNIV_PFS_IO struct PSI_file_locker* locker = NULL; PSI_file_locker_state state; - register_pfs_file_io_begin(&state, locker, - fd, 0, PSI_FILE_CLOSE, - __FILE__, __LINE__); + locker = PSI_FILE_CALL(get_thread_file_descriptor_locker)( + &state, fd, PSI_FILE_CLOSE); + if (locker != NULL) { + PSI_FILE_CALL(start_file_wait)( + locker, 0, __FILE__, __LINE__); + } #endif if (fd >= 0) { close(fd); } #ifdef UNIV_PFS_IO - register_pfs_file_io_end(locker, 0); + if (locker != NULL) { + PSI_FILE_CALL(end_file_wait)(locker, 0); + } #endif } /*********************************************************************//** diff --git a/storage/innobase/row/row0sel.cc b/storage/innobase/row/row0sel.cc index 42b1ab57630..c53c1154f60 100644 --- a/storage/innobase/row/row0sel.cc +++ b/storage/innobase/row/row0sel.cc @@ -61,6 +61,9 @@ Created 12/19/1997 Heikki Tuuri #include "my_sys.h" /* DEBUG_SYNC_C */ #include "my_compare.h" /* enum icp_result */ +#include "thr_lock.h" +#include "handler.h" +#include "ha_innodb.h" /* Maximum number of rows to prefetch; MySQL interface has another parameter */ #define SEL_MAX_N_PREFETCH 16 @@ -2743,34 +2746,43 @@ row_sel_field_store_in_mysql_format_func( #ifdef UNIV_DEBUG /** Convert a field from Innobase format to MySQL format. */ -# define row_sel_store_mysql_field(m,p,r,i,o,f,t) \ - row_sel_store_mysql_field_func(m,p,r,i,o,f,t) +# define row_sel_store_mysql_field(m,p,r,i,o,f,t,c) \ + row_sel_store_mysql_field_func(m,p,r,i,o,f,t,c) #else /* UNIV_DEBUG */ /** Convert a field from Innobase format to MySQL format. */ -# define row_sel_store_mysql_field(m,p,r,i,o,f,t) \ - row_sel_store_mysql_field_func(m,p,r,o,f,t) +# define row_sel_store_mysql_field(m,p,r,i,o,f,t,c) \ + row_sel_store_mysql_field_func(m,p,r,o,f,t,c) #endif /* UNIV_DEBUG */ -/**************************************************************//** -Convert a field in the Innobase format to a field in the MySQL format. */ +/** Convert a field in the Innobase format to a field in the MySQL format. +@param[out] mysql_rec record in the MySQL format +@param[in,out] prebuilt prebuilt struct +@param[in] rec InnoDB record; must be protected + by a page latch +@param[in] index index of rec +@param[in] offsets array returned by rec_get_offsets() +@param[in] field_no templ->rec_field_no or + templ->clust_rec_field_no + or templ->icp_rec_field_no + or sec field no if clust_templ_for_sec + is TRUE +@param[in] templ row template +@param[in] clust_templ_for_sec TRUE if rec belongs to secondary index + but prebuilt template is in clustered + index format and used only for end + range comparison. */ static MY_ATTRIBUTE((warn_unused_result)) ibool row_sel_store_mysql_field_func( -/*===========================*/ - byte* mysql_rec, /*!< out: record in the - MySQL format */ - row_prebuilt_t* prebuilt, /*!< in/out: prebuilt struct */ - const rec_t* rec, /*!< in: InnoDB record; - must be protected by - a page latch */ + byte* mysql_rec, + row_prebuilt_t* prebuilt, + const rec_t* rec, #ifdef UNIV_DEBUG - const dict_index_t* index, /*!< in: index of rec */ + const dict_index_t* index, #endif - const ulint* offsets, /*!< in: array returned by - rec_get_offsets() */ - ulint field_no, /*!< in: templ->rec_field_no or - templ->clust_rec_field_no or - templ->icp_rec_field_no */ - const mysql_row_templ_t*templ) /*!< in: row template */ + const ulint* offsets, + ulint field_no, + const mysql_row_templ_t*templ, + bool clust_templ_for_sec) { const byte* data; ulint len; @@ -2779,10 +2791,12 @@ row_sel_store_mysql_field_func( ut_ad(templ); ut_ad(templ >= prebuilt->mysql_template); ut_ad(templ < &prebuilt->mysql_template[prebuilt->n_template]); - ut_ad(field_no == templ->clust_rec_field_no + ut_ad(clust_templ_for_sec + || field_no == templ->clust_rec_field_no || field_no == templ->rec_field_no || field_no == templ->icp_rec_field_no); - ut_ad(rec_offs_validate(rec, index, offsets)); + ut_ad(rec_offs_validate(rec, + clust_templ_for_sec == true ? prebuilt->index : index, offsets)); if (UNIV_UNLIKELY(rec_offs_nth_extern(offsets, field_no))) { @@ -2896,30 +2910,37 @@ row_sel_store_mysql_field_func( return(TRUE); } -/**************************************************************//** -Convert a row in the Innobase format to a row in the MySQL format. +/** Convert a row in the Innobase format to a row in the MySQL format. Note that the template in prebuilt may advise us to copy only a few columns to mysql_rec, other columns are left blank. All columns may not be needed in the query. +@param[out] mysql_rec row in the MySQL format +@param[in] prebuilt prebuilt structure +@param[in] rec Innobase record in the index + which was described in prebuilt's + template, or in the clustered index; + must be protected by a page latch +@param[in] rec_clust TRUE if the rec in the clustered index +@param[in] index index of rec +@param[in] offsets array returned by rec_get_offsets(rec) +@param[in] clust_templ_for_sec TRUE if rec belongs to secondary index + but the prebuilt->template is in + clustered index format and it is + used only for end range comparison @return TRUE on success, FALSE if not all columns could be retrieved */ static MY_ATTRIBUTE((warn_unused_result)) ibool row_sel_store_mysql_rec( -/*====================*/ - byte* mysql_rec, /*!< out: row in the MySQL format */ - row_prebuilt_t* prebuilt, /*!< in: prebuilt struct */ - const rec_t* rec, /*!< in: Innobase record in the index - which was described in prebuilt's - template, or in the clustered index; - must be protected by a page latch */ - ibool rec_clust, /*!< in: TRUE if rec is in the - clustered index instead of - prebuilt->index */ - const dict_index_t* index, /*!< in: index of rec */ - const ulint* offsets) /*!< in: array returned by - rec_get_offsets(rec) */ + byte* mysql_rec, + row_prebuilt_t* prebuilt, + const rec_t* rec, + ibool rec_clust, + const dict_index_t* index, + const ulint* offsets, + bool clust_templ_for_sec) { - ulint i; + ulint i; + std::vector<ulint> template_col; ut_ad(rec_clust || index == prebuilt->index); ut_ad(!rec_clust || dict_index_is_clust(index)); @@ -2929,9 +2950,20 @@ row_sel_store_mysql_rec( prebuilt->blob_heap = NULL; } + if (clust_templ_for_sec) { + /* Store all clustered index field of + secondary index record. */ + for (i = 0; i < dict_index_get_n_fields( + prebuilt->index); i++) { + ulint sec_field = dict_index_get_nth_field_pos( + index, prebuilt->index, i); + template_col.push_back(sec_field); + } + } + for (i = 0; i < prebuilt->n_template; i++) { const mysql_row_templ_t*templ = &prebuilt->mysql_template[i]; - const ulint field_no + ulint field_no = rec_clust ? templ->clust_rec_field_no : templ->rec_field_no; @@ -2940,9 +2972,24 @@ row_sel_store_mysql_rec( ut_ad(dict_index_get_nth_field(index, field_no)->prefix_len == 0); + if (clust_templ_for_sec) { + std::vector<ulint>::iterator it; + it = std::find(template_col.begin(), + template_col.end(), field_no); + + if (it == template_col.end()) { + continue; + } + + ut_ad(templ->rec_field_no == templ->clust_rec_field_no); + + field_no = it - template_col.begin(); + } + if (!row_sel_store_mysql_field(mysql_rec, prebuilt, rec, index, offsets, - field_no, templ)) { + field_no, templ, + clust_templ_for_sec)) { return(FALSE); } } @@ -3582,7 +3629,7 @@ row_search_idx_cond_check( if (!row_sel_store_mysql_field(mysql_rec, prebuilt, rec, prebuilt->index, offsets, templ->icp_rec_field_no, - templ)) { + templ, false)) { return(ICP_NO_MATCH); } } @@ -3603,7 +3650,7 @@ row_search_idx_cond_check( || dict_index_is_clust(prebuilt->index)) { if (!row_sel_store_mysql_rec( mysql_rec, prebuilt, rec, FALSE, - prebuilt->index, offsets)) { + prebuilt->index, offsets, false)) { ut_ad(dict_index_is_clust(prebuilt->index)); return(ICP_NO_MATCH); } @@ -3622,6 +3669,27 @@ row_search_idx_cond_check( return(result); } +/** Check the pushed down end range condition to avoid extra traversal +if records are not within view and also to avoid prefetching in the +cache buffer. +@param[in] mysql_rec record in MySQL format +@param[in,out] handler the MySQL handler performing the scan +@retval true if the row in mysql_rec is out of range +@retval false if the row in mysql_rec is in range */ +static +bool +row_search_end_range_check( + const byte* mysql_rec, + ha_innobase* handler) +{ + if (handler->end_range && + handler->compare_key_in_buffer(mysql_rec) > 0) { + return(true); + } + + return(false); +} + /********************************************************************//** Searches for rows in the database. This is used in the interface to MySQL. This function opens a cursor, and also implements fetch next @@ -3659,7 +3727,9 @@ row_search_for_mysql( trx_t* trx = prebuilt->trx; dict_index_t* clust_index; que_thr_t* thr; - const rec_t* rec; + const rec_t* prev_rec = NULL; + const rec_t* rec = NULL; + byte* end_range_cache = NULL; const rec_t* result_rec = NULL; const rec_t* clust_rec; dberr_t err = DB_SUCCESS; @@ -3684,6 +3754,7 @@ row_search_for_mysql( ulint* offsets = offsets_; ibool table_lock_waited = FALSE; byte* next_buf = 0; + ulint end_loop = 0; rec_offs_init(offsets_); @@ -3823,6 +3894,10 @@ row_search_for_mysql( err = DB_SUCCESS; goto func_exit; + } else if (prebuilt->end_range == true) { + prebuilt->end_range = false; + err = DB_RECORD_NOT_FOUND; + goto func_exit; } if (prebuilt->fetch_cache_first > 0 @@ -3956,7 +4031,8 @@ row_search_for_mysql( if (!row_sel_store_mysql_rec( buf, prebuilt, - rec, FALSE, index, offsets)) { + rec, FALSE, index, + offsets, false)) { /* Only fresh inserts may contain incomplete externally stored columns. Pretend that such @@ -4205,11 +4281,62 @@ rec_loop: and neither can a record lock be placed on it: we skip such a record. */ + prev_rec = NULL; goto next_rec; } if (page_rec_is_supremum(rec)) { + /** Compare the last record of the page with end range + passed to InnoDB when there is no ICP and number of loops + in row_search_for_mysql for rows found but not + reporting due to search views etc. */ + if (prev_rec != NULL + && prebuilt->mysql_handler->end_range != NULL + && prebuilt->idx_cond == NULL + && end_loop >= 100) { + + dict_index_t* key_index = prebuilt->index; + bool clust_templ_for_sec = false; + + if (end_range_cache == NULL) { + end_range_cache = static_cast<byte*>( + ut_malloc(prebuilt->mysql_row_len)); + } + + if (index != clust_index + && prebuilt->need_to_access_clustered) { + /** Secondary index record but the template + based on PK. */ + key_index = clust_index; + clust_templ_for_sec = true; + } + + /** Create offsets based on prebuilt index. */ + offsets = rec_get_offsets(prev_rec, prebuilt->index, + offsets, ULINT_UNDEFINED, &heap); + + if (row_sel_store_mysql_rec( + end_range_cache, prebuilt, prev_rec, + clust_templ_for_sec, key_index, offsets, + clust_templ_for_sec)) { + + if (row_search_end_range_check( + end_range_cache, + prebuilt->mysql_handler)) { + + /** In case of prebuilt->fetch, + set the error in prebuilt->end_range. */ + if (prebuilt->n_fetch_cached > 0) { + prebuilt->end_range = true; + } + + err = DB_RECORD_NOT_FOUND; + goto normal_return; + } + } + } + if (set_also_gap_locks && !(srv_locks_unsafe_for_binlog || trx->isolation_level <= TRX_ISO_READ_COMMITTED) @@ -4241,6 +4368,7 @@ rec_loop: /* A page supremum record cannot be in the result set: skip it now that we have placed a possible lock on it */ + prev_rec = NULL; goto next_rec; } @@ -4308,6 +4436,7 @@ wrong_offs: btr_pcur_move_to_last_on_page(pcur, &mtr); + prev_rec = NULL; goto next_rec; } } @@ -4334,10 +4463,13 @@ wrong_offs: fputs(". We try to skip the record.\n", stderr); + prev_rec = NULL; goto next_rec; } } + prev_rec = rec; + /* Note that we cannot trust the up_match value in the cursor at this place because we can arrive here after moving the cursor! Thus we have to recompare rec and search_tuple to determine if they @@ -4562,6 +4694,7 @@ no_gap_lock: did_semi_consistent_read = TRUE; rec = old_vers; + prev_rec = rec; break; default: @@ -4608,6 +4741,7 @@ no_gap_lock: } rec = old_vers; + prev_rec = rec; } } else { /* We are looking into a non-clustered index, @@ -4785,7 +4919,7 @@ requires_clust_rec: appropriate version of the clustered index record. */ if (!row_sel_store_mysql_rec( buf, prebuilt, result_rec, - TRUE, clust_index, offsets)) { + TRUE, clust_index, offsets, false)) { goto next_rec; } } @@ -4853,7 +4987,7 @@ requires_clust_rec: next_buf, prebuilt, result_rec, result_rec != rec, result_rec != rec ? clust_index : index, - offsets)) { + offsets, false)) { if (next_buf == buf) { ut_a(prebuilt->n_fetch_cached == 0); @@ -4908,7 +5042,7 @@ requires_clust_rec: buf, prebuilt, result_rec, result_rec != rec, result_rec != rec ? clust_index : index, - offsets)) { + offsets, false)) { /* Only fresh inserts may contain incomplete externally stored columns. Pretend that such records do @@ -4960,6 +5094,8 @@ idx_cond_failed: goto normal_return; next_rec: + end_loop++; + /* Reset the old and new "did semi-consistent read" flags. */ if (UNIV_UNLIKELY(prebuilt->row_read_type == ROW_READ_DID_SEMI_CONSISTENT)) { @@ -5146,6 +5282,11 @@ normal_return: func_exit: trx->op_info = ""; + + if (end_range_cache != NULL) { + ut_free(end_range_cache); + } + if (UNIV_LIKELY_NULL(heap)) { mem_heap_free(heap); } diff --git a/storage/innobase/srv/srv0srv.cc b/storage/innobase/srv/srv0srv.cc index a67f3a776c5..173577740f8 100644 --- a/storage/innobase/srv/srv0srv.cc +++ b/storage/innobase/srv/srv0srv.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1995, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1995, 2017, Oracle and/or its affiliates. All Rights Reserved. Copyright (c) 2008, 2009 Google Inc. Copyright (c) 2009, Percona Inc. @@ -2315,6 +2315,8 @@ DECLARE_THREAD(srv_master_thread)( /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); + srv_slot_t* slot; ulint old_activity_count = srv_get_activity_count(); ib_time_t last_print_time; @@ -2376,6 +2378,7 @@ suspend_thread: os_event_wait(slot->event); if (srv_shutdown_state == SRV_SHUTDOWN_EXIT_THREADS) { + my_thread_end(); os_thread_exit(NULL); } @@ -2458,6 +2461,8 @@ DECLARE_THREAD(srv_worker_thread)( void* arg MY_ATTRIBUTE((unused))) /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); + srv_slot_t* slot; ut_ad(!srv_read_only_mode); @@ -2516,6 +2521,7 @@ DECLARE_THREAD(srv_worker_thread)( os_thread_pf(os_thread_get_curr_id())); #endif /* UNIV_DEBUG_THREAD_CREATION */ + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ os_thread_exit(NULL); @@ -2716,6 +2722,8 @@ DECLARE_THREAD(srv_purge_coordinator_thread)( void* arg MY_ATTRIBUTE((unused))) /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); + srv_slot_t* slot; ulint n_total_purged = ULINT_UNDEFINED; @@ -2821,6 +2829,7 @@ DECLARE_THREAD(srv_purge_coordinator_thread)( srv_release_threads(SRV_WORKER, srv_n_purge_threads - 1); } + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ os_thread_exit(NULL); @@ -2889,3 +2898,18 @@ srv_purge_wakeup(void) } } +/** Check whether given space id is undo tablespace id +@param[in] space_id space id to check +@return true if it is undo tablespace else false. */ +bool +srv_is_undo_tablespace( + ulint space_id) +{ + if (srv_undo_space_id_start == 0) { + return (false); + } + + return(space_id >= srv_undo_space_id_start + && space_id < (srv_undo_space_id_start + + srv_undo_tablespaces_open)); +} diff --git a/storage/innobase/srv/srv0start.cc b/storage/innobase/srv/srv0start.cc index 697107a1e0d..f6b68a73b74 100644 --- a/storage/innobase/srv/srv0start.cc +++ b/storage/innobase/srv/srv0start.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2016, Oracle and/or its affiliates. All rights reserved. +Copyright (c) 1996, 2017, Oracle and/or its affiliates. All rights reserved. Copyright (c) 2008, Google Inc. Copyright (c) 2009, Percona Inc. @@ -109,6 +109,9 @@ UNIV_INTERN ibool srv_have_fullfsync = FALSE; /** TRUE if a raw partition is in use */ UNIV_INTERN ibool srv_start_raw_disk_in_use = FALSE; +/** UNDO tablespaces starts with space id. */ +ulint srv_undo_space_id_start; + /** TRUE if the server is being started, before rolling back any incomplete transactions */ UNIV_INTERN ibool srv_startup_is_before_trx_rollback_phase = FALSE; @@ -124,7 +127,7 @@ SRV_SHUTDOWN_CLEANUP and then to SRV_SHUTDOWN_LAST_PHASE, and so on */ UNIV_INTERN enum srv_shutdown_state srv_shutdown_state = SRV_SHUTDOWN_NONE; /** Files comprising the system tablespace */ -static os_file_t files[1000]; +static pfs_os_file_t files[1000]; /** io_handler_thread parameters for thread identification */ static ulint n[SRV_MAX_N_IO_THREADS + 6]; @@ -530,7 +533,7 @@ static MY_ATTRIBUTE((nonnull, warn_unused_result)) dberr_t create_log_file( /*============*/ - os_file_t* file, /*!< out: file handle */ + pfs_os_file_t* file, /*!< out: file handle */ const char* name) /*!< in: log file name */ { ibool ret; @@ -737,7 +740,7 @@ static MY_ATTRIBUTE((nonnull, warn_unused_result)) dberr_t open_log_file( /*==========*/ - os_file_t* file, /*!< out: file handle */ + pfs_os_file_t* file, /*!< out: file handle */ const char* name, /*!< in: log file name */ os_offset_t* size) /*!< out: file size */ { @@ -853,7 +856,7 @@ open_or_create_data_files( && os_file_get_last_error(false) != OS_FILE_ALREADY_EXISTS #ifdef UNIV_AIX - /* AIX 5.1 after security patch ML7 may have + /* AIX 5.1 after security patch ML7 may have errno set to 0 here, which causes our function to return 100; work around that AIX problem */ @@ -1156,7 +1159,7 @@ srv_undo_tablespace_create( const char* name, /*!< in: tablespace name */ ulint size) /*!< in: tablespace size in pages */ { - os_file_t fh; + pfs_os_file_t fh; ibool ret; dberr_t err = DB_SUCCESS; @@ -1233,7 +1236,7 @@ srv_undo_tablespace_open( const char* name, /*!< in: tablespace name */ ulint space) /*!< in: tablespace id */ { - os_file_t fh; + pfs_os_file_t fh; dberr_t err = DB_ERROR; ibool ret; ulint flags; @@ -1332,13 +1335,23 @@ srv_undo_tablespaces_init( for (i = 0; create_new_db && i < n_conf_tablespaces; ++i) { char name[OS_FILE_MAX_PATH]; + ulint space_id = i + 1; + + DBUG_EXECUTE_IF("innodb_undo_upgrade", + space_id = i + 3;); ut_snprintf( name, sizeof(name), "%s%cundo%03lu", - srv_undo_dir, SRV_PATH_SEPARATOR, i + 1); + srv_undo_dir, SRV_PATH_SEPARATOR, space_id); + + if (i == 0) { + srv_undo_space_id_start = space_id; + prev_space_id = srv_undo_space_id_start - 1; + } + + undo_tablespace_ids[i] = space_id; - /* Undo space ids start from 1. */ err = srv_undo_tablespace_create( name, SRV_UNDO_TABLESPACE_SIZE_IN_PAGES); @@ -1360,14 +1373,16 @@ srv_undo_tablespaces_init( if (!create_new_db) { n_undo_tablespaces = trx_rseg_get_n_undo_tablespaces( undo_tablespace_ids); - } else { - n_undo_tablespaces = n_conf_tablespaces; - for (i = 1; i <= n_undo_tablespaces; ++i) { - undo_tablespace_ids[i - 1] = i; + if (n_undo_tablespaces != 0) { + srv_undo_space_id_start = undo_tablespace_ids[0]; + prev_space_id = srv_undo_space_id_start - 1; } - undo_tablespace_ids[i] = ULINT_UNDEFINED; + } else { + n_undo_tablespaces = n_conf_tablespaces; + + undo_tablespace_ids[n_conf_tablespaces] = ULINT_UNDEFINED; } /* Open all the undo tablespaces that are currently in use. If we @@ -1391,8 +1406,6 @@ srv_undo_tablespaces_init( ut_a(undo_tablespace_ids[i] != 0); ut_a(undo_tablespace_ids[i] != ULINT_UNDEFINED); - /* Undo space ids start from 1. */ - err = srv_undo_tablespace_open(name, undo_tablespace_ids[i]); if (err != DB_SUCCESS) { @@ -1427,11 +1440,23 @@ srv_undo_tablespaces_init( break; } + /** Note the first undo tablespace id in case of + no active undo tablespace. */ + if (n_undo_tablespaces == 0) { + srv_undo_space_id_start = i; + } + ++n_undo_tablespaces; ++*n_opened; } + /** Explictly specify the srv_undo_space_id_start + as zero when there are no undo tablespaces. */ + if (n_undo_tablespaces == 0) { + srv_undo_space_id_start = 0; + } + /* If the user says that there are fewer than what we find we tolerate that discrepancy but not the inverse. Because there could be unused undo tablespaces for future use. */ @@ -1476,10 +1501,11 @@ srv_undo_tablespaces_init( mtr_start(&mtr); /* The undo log tablespace */ - for (i = 1; i <= n_undo_tablespaces; ++i) { + for (i = 0; i < n_undo_tablespaces; ++i) { fsp_header_init( - i, SRV_UNDO_TABLESPACE_SIZE_IN_PAGES, &mtr); + undo_tablespace_ids[i], + SRV_UNDO_TABLESPACE_SIZE_IN_PAGES, &mtr); } mtr_commit(&mtr); @@ -1554,6 +1580,10 @@ innobase_start_or_create_for_mysql(void) char* logfile0 = NULL; size_t dirnamelen; + if (srv_force_recovery == SRV_FORCE_NO_LOG_REDO) { + srv_read_only_mode = 1; + } + high_level_read_only = srv_read_only_mode || srv_force_recovery > SRV_FORCE_NO_TRX_UNDO; diff --git a/storage/innobase/trx/trx0purge.cc b/storage/innobase/trx/trx0purge.cc index efc600d16b1..55af54991c6 100644 --- a/storage/innobase/trx/trx0purge.cc +++ b/storage/innobase/trx/trx0purge.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -692,7 +692,8 @@ trx_purge_get_rseg_with_min_trx_id( /* We assume in purge of externally stored fields that space id is in the range of UNDO tablespace space ids */ - ut_a(purge_sys->rseg->space <= srv_undo_tablespaces_open); + ut_a(purge_sys->rseg->space == 0 + || srv_is_undo_tablespace(purge_sys->rseg->space)); zip_size = purge_sys->rseg->zip_size; diff --git a/storage/innobase/trx/trx0roll.cc b/storage/innobase/trx/trx0roll.cc index e1e253cbb76..09e8e018c4f 100644 --- a/storage/innobase/trx/trx0roll.cc +++ b/storage/innobase/trx/trx0roll.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2016, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -800,6 +800,7 @@ DECLARE_THREAD(trx_rollback_or_clean_all_recovered)( /*!< in: a dummy parameter required by os_thread_create */ { + my_thread_init(); ut_ad(!srv_read_only_mode); #ifdef UNIV_PFS_THREAD @@ -810,6 +811,7 @@ DECLARE_THREAD(trx_rollback_or_clean_all_recovered)( trx_rollback_or_clean_is_active = false; + my_thread_end(); /* We count the number of threads in os_thread_exit(). A created thread should always use that to exit and not use return() to exit. */ diff --git a/storage/innobase/trx/trx0sys.cc b/storage/innobase/trx/trx0sys.cc index e5f03f4b96a..00d79e7ff51 100644 --- a/storage/innobase/trx/trx0sys.cc +++ b/storage/innobase/trx/trx0sys.cc @@ -1,6 +1,6 @@ /***************************************************************************** -Copyright (c) 1996, 2015, Oracle and/or its affiliates. All Rights Reserved. +Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software @@ -909,18 +909,12 @@ trx_sys_create_rsegs( ulint new_rsegs = n_rsegs - n_used; for (i = 0; i < new_rsegs; ++i) { - ulint space; + ulint space_id; + space_id = (n_spaces == 0) ? 0 + : (srv_undo_space_id_start + i % n_spaces); - /* Tablespace 0 is the system tablespace. All UNDO - log tablespaces start from 1. */ - - if (n_spaces > 0) { - space = (i % n_spaces) + 1; - } else { - space = 0; /* System tablespace */ - } - - if (trx_rseg_create(space) != NULL) { + /* Tablespace 0 is the system tablespace. */ + if (trx_rseg_create(space_id) != NULL) { ++n_used; } else { break; |